figma-ai-project

design MCP Server

Generates production-ready Mantine v7 components from Figma designs using a 3-stage AI workflow (Plan → Act → Reflect), validated by Storybook and Playwright.

VerifiedInstall Ready
designdesign
9 views0 stars0 forksv1.0.0GPL-3.0

Why This Matters

Discovered via github-topic:mcp-server and last synced 3mo ago.

VerifiedInstall Ready
Source
github-topic:mcp-server
Stars
0
Last synced
3mo ago
Install
Instructions detected

Install

1. Install the package

npm install figma-ai-project
29
Tools
0
Resources
0
Prompts
Standard I/O
Transport

Available Tools (29)

mantine_status

Poll a running job — returns state, timestamps, recent log tail.

10

Visual Snapshot

claude-opus-4-5

Break-glass only — fires if escalation Step 1 also fails; ~15× Haiku cost

mantine_get_plan

Read the full plan markdown after `mantine_plan` completes.

claude-haiku-4-5-20251001

Rewrites dry `detail` fields into assertive architect persona before posting to Figma

Gate

What it checks

7

Size Variant Coverage

8

Test Coverage

mantine_get_files

Read all 4 generated files after `mantine_generate` completes.

5

Tracker Update

Layer

Package

6

data-* Attributes

Icons

`@tabler/icons-react`

mantine_generate

Stage 2+3 — write TSX/CSS/stories/spec. Skips Stage 1 if a plan exists. ~15–20 min.

9

Dependency Audit

File

Stage

Requirement

Notes

2

File Integrity

Stage

Model

Testing

`@playwright/test`

Tool

Description

3

PostCSS Standard

postcss-preset-mantine

—

1

Token Compliance

11

Portal CSS Variable Scope

Storybook

`@storybook/react-vite`

4

Storybook Autodocs

claude-sonnet-4-6

`--escalate` flag: fresh Sonnet run after the 3-iteration self-healing cycle fails

mantine_plan

Stage 1 — query Figma, produce architectural plan. ~5–10 min. Returns instantly on cache hit (<24 h).